""" App Controller Agent — main_agent.py ===================================== Opens, closes, focuses, and lists applications via natural language. Usage: run_agent("open VS Code") run_agent("close Postman") run_agent("switch to Chrome") run_agent("what apps are open?") run_agent("open VS Code in ~/Documents/AI OS") run_agent("open Chrome at notion.so") """ import os import json from groq import Groq from dotenv import load_dotenv load_dotenv(os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", ".env")) from app_launcher import open_app, close_app, focus_app, list_open_apps, REGISTRY, get_vscode_recent_folders # ── Agent metadata ──────────────────────────────────────────────────────────── AGENT_META = { "name": "app_controller", "description": "Opens, closes, and switches focus between applications on the desktop.", "capabilities": [ "Open any installed application by name", "Open VS Code or terminal in a specific folder", "Open Chrome or Firefox at a specific URL", "Close any running application", "Switch focus to any open application", "List all currently open applications", ], "call_when": [ "User says open, launch, start followed by an app name", "User says 'open vscode', 'open chrome', 'open terminal'", "User says close, quit, kill followed by an app name", "User says switch to, focus, bring up followed by an app name", "User asks what apps are open, what's running", "User says open VS Code in a folder", "User says open Chrome at a URL", ], "does_not_own": [ "Installing applications", "System settings or configuration", "File operations beyond opening in an app", "Creating files or writing code inside VS Code" ], "chains_with": [], "entry": "run_agent", } # ── LLM command parser ──────────────────────────────────────────────────────── PARSER_PROMPT = """You are a command parser for a Linux desktop app controller. Parse the user's message into a structured command. AVAILABLE ACTIONS: - open → launch an application (optionally with path or url) - close → close/quit an application - focus → switch focus to an already-open application - list → list all currently open applications KNOWN APPS (use exact key names): {app_list} RULES: - "open", "launch", "start", "run" → action: open - "close", "quit", "kill", "exit", "shut down" → action: close - "switch to", "focus", "bring up", "go to", "show" → action: focus - "what's open", "what apps", "what's running", "list apps" → action: list - If user says "open VS Code in ~/Documents/AI OS" → action: open, app: vscode, path: ~/Documents/AI OS - If user says "open Chrome at notion.so" → action: open, app: chrome, url: https://notion.so - If app is not in known list, use the raw name as-is in app field - For list action, app field should be empty string Respond ONLY with valid JSON, no markdown: { "action": "open|close|focus|list", "app": "", "path": "", "url": "", "confidence": 0.0-1.0, "reasoning": "" }""" def _parse_command(query: str) -> dict: """Parse directly — no LLM needed for simple app commands.""" import re q = query.lower().strip() # ── Action ──────────────────────────────────────────────────────────── if any(w in q for w in ["close", "quit", "kill", "exit", "shut"]): action = "close" elif any(w in q for w in ["switch to", "focus", "bring up", "show", "go to"]): action = "focus" elif any(w in q for w in ["list", "what's open", "what apps", "running", "what is open"]): action = "list" else: action = "open" # ── Path — extract "in " or "at " ───────────────────────── path = "" path_match = re.search(r"\b(?:in|at)\s+([~/][\S]*)", query) if path_match: path = path_match.group(1) # ── App — remove action words + path phrase ─────────────────────────── app = q for phrase in ["open", "launch", "start", "close", "quit", "kill", "switch to", "focus", "bring up", "please", "can you", "hey"]: app = app.replace(phrase, " ") if path: app = app.replace(f"in {path.lower()}", "").replace(f"at {path.lower()}", "") app = " ".join(app.split()) # normalize whitespace # ── Resolve app alias ───────────────────────────────────────────────── from app_launcher import resolve_app key, _ = resolve_app(app) if key: app = key return { "action": action, "app": app.strip(), "path": path, "url": "", "confidence": 0.95, "reasoning": "direct parse", } NEW_WINDOW_PHRASES = ["new window", "another window", "second window", "open again", "reopen"] def _fallback_parse(query: str) -> dict: q = query.lower() action = "open" if any(w in q for w in ["close", "quit", "kill", "exit", "shut"]): action = "close" elif any(w in q for w in ["switch", "focus", "bring", "show", "go to"]): action = "focus" elif any(w in q for w in ["list", "what's open", "what apps", "running"]): action = "list" wants_new = any(p in q for p in NEW_WINDOW_PHRASES) import re as _re path = "" path_match = _re.search(r"\bin\s+([\S]+)", query) if path_match: path = path_match.group(1) app = q for word in ["open", "close", "launch", "start", "quit", "kill", "switch to", "focus", "bring up", "please", "can you", "hey", "in " + path]: # strip "in /path" too app = app.replace(word, "").replace(word.lower(), "").strip() return {"action": action, "app": app.strip(), "path": "", "url": "", "force_new": wants_new, # ← NEW "confidence": 0.6, "reasoning": "fallback"} # ── Response formatter ──────────────────────────────────────────────────────── def _format_response(action: str, result: dict) -> str: """Format the launcher result into a clean chat response.""" if action == "list": apps = result.get("apps", []) if not apps: return "No applications are currently open." known = [a["name"] for a in apps if a["known"]] unknown = [a["title"] for a in apps if not a["known"]] lines = [] if known: lines.append("**Open apps:** " + ", ".join(known)) if unknown: lines.append("**Other windows:** " + ", ".join(unknown[:5])) return "\n".join(lines) if result.get("action") == "already_open": app_key = result.get("app", "") return f"__CONFIRM_NEW_WINDOW__{app_key}" return result.get("message", "Done.") # ── Main entry point ────────────────────────────────────────────────────────── def run_agent(query: str = "", **kwargs) -> str: """ Parse natural language command and execute it. Args: query: Natural language command e.g. "open VS Code", "close Postman" Returns: Human-readable result string. """ force_new = query.startswith("__FORCE_NEW__") or kwargs.get("force_new", False) pending_path = kwargs.get("pending_path", "") create_folder = kwargs.get("create_folder", False) if pending_path: if not pending_path.startswith("~") and not pending_path.startswith("/"): pending_path = "~/" + pending_path.lstrip("/") expanded = os.path.expanduser(pending_path) print(f"[AppController] pending_path={pending_path} expanded={expanded}") print(f"[AppController] exists={os.path.exists(expanded)} isdir={os.path.isdir(expanded)}") try: os.makedirs(expanded, exist_ok=True) print(f"[AppController] makedirs done, isdir now={os.path.isdir(expanded)}") except Exception as e: print(f"[AppController] makedirs FAILED: {e}") result = open_app("vscode", path=expanded) return _format_response("open", result) if query.startswith("__FORCE_NEW__"): query = query.replace("__FORCE_NEW__", "", 1) print(f"\n[AppController] Query: {query}, force_new: {force_new}") if not query.strip(): return "What would you like me to open, close, or switch to?" command = _parse_command(query) action = command.get("action", "open") app = command.get("app", "") path = command.get("path", "") url = command.get("url", "") print(f"[AppController] → {action} '{app}'" + (f" path={path}" if path else "") + (f" url={url}" if url else "")) if action == "open": VSCODE_KEYS = ["vscode", "vs code", "visual studio code", "code"] is_vscode = any(k in app.lower() for k in VSCODE_KEYS) # No path — ask user to pick/create folder if is_vscode and not path: recent = [] try: from app_launcher import get_vscode_recent_folders recent = get_vscode_recent_folders() except Exception: pass import json return f"__ASK_VSCODE_PATH__{json.dumps(recent)}" # Path provided in query — ensure it's a folder if is_vscode and path: # Fix missing home prefix if path.startswith("/Documents") or path.startswith("/Downloads") or path.startswith("/Desktop"): path = os.path.expanduser("~") + path elif not path.startswith("~") and not path.startswith("/"): path = "~/" + path.lstrip("/") expanded = os.path.expanduser(path) try: os.makedirs(expanded, exist_ok=True) print(f"[AppController] Ensured folder: {expanded}") except PermissionError: return f"Cannot create folder at {path} — permission denied. Try ~/Documents/test1234 instead." except Exception as e: return f"Could not create folder: {e}" path = expanded result = open_app(app, path=path, url=url, force_new=force_new) if result.get("action") == "already_open": return f"__CONFIRM_NEW_WINDOW__{result.get('app', app)}" elif action == "close": result = close_app(app) elif action == "focus": result = focus_app(app) elif action == "list": result = list_open_apps() else: result = {"success": False, "message": f"Unknown action: {action}"} return _format_response(action, result) # ── CLI test ────────────────────────────────────────────────────────────────── if __name__ == "__main__": import sys query = " ".join(sys.argv[1:]) if len(sys.argv) > 1 else "what apps are open" print(run_agent(query))